Assignment 03

Author

Affiliation

Ryan Martin

Boston University

from pyspark.sql import SparkSession
import pandas as pd
import plotly.express as px
import plotly.io as pio
pio.renderers.default = "svg"
import re
import numpy as np
import plotly.graph_objects as go
from pyspark.sql.functions import col, split, explode, regexp_replace, transform, when
from pyspark.sql import functions as F
from pyspark.sql.functions import col, monotonically_increasing_id

np.random.seed(42)

pio.renderers.default = "notebook"

# Initialize Spark Session
spark = SparkSession.builder.appName("./data/LightcastData").getOrCreate()

# Load Data
df = spark.read.option("header", "true").option("inferSchema", "true").option("multiLine","true").option("escape", "\"").csv("./data/lightcast_job_postings.csv")

# Show Schema and Sample Data
#print("---This is Diagnostic check, No need to print it in the final doc---")

#df.printSchema() # comment this line when rendering the submission
#df.show(5)

WARNING: Using incubator modules: jdk.incubator.vector
Using Spark's default log4j profile: org/apache/spark/log4j2-defaults.properties
Setting default log level to "WARN".
To adjust logging level use sc.setLogLevel(newLevel). For SparkR, use setLogLevel(newLevel).
25/09/23 23:53:53 WARN NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
[Stage 1:>                                                          (0 + 1) / 1]                                                                                

1 1. Data Preperation

df = df.withColumn("SALARY", col("SALARY").cast("float")) \
      .withColumn("SALARY_FROM", col("SALARY_FROM").cast("float")) \
      .withColumn("SALARY_TO", col("SALARY_TO").cast("float")) \
      .withColumn("MIN_YEARS_EXPERIENCE", col("MIN_YEARS_EXPERIENCE").cast("float")) \
      .withColumn("MAX_YEARS_EXPERIENCE", col("MAX_YEARS_EXPERIENCE").cast("float")) \
      
def compute_median(sdf, col_name):
  q = sdf.approxQuantile(col_name, [0.5], 0.01)
  return q[0] if q else None

median_from = compute_median(df, "SALARY_FROM")
median_to = compute_median(df, "SALARY_TO")
median_salary = compute_median(df, "SALARY")

print("Medians:", median_from, median_to, median_salary)

df = df.fillna({
      "SALARY_FROM": median_from,
      "SALARY_TO": median_to,
      "SALARY": median_salary
})

df= df.withColumn("Average_Salary", (col("SALARY_FROM") + col("SALARY_TO")) / 2)


export_cols = [
    "EDUCATION_LEVELS_NAME",
    "REMOTE_TYPE_NAME",
    "MAX_YEARS_EXPERIENCE",
    "Average_Salary",
    "SALARY",
    "LOT_V6_SPECIALIZED_OCCUPATION_NAME"
] 

df_selected = df.select(*export_cols)

pdf = df_selected.toPandas()
pdf.to_csv("./data/lightcast_cleaned.csv", index=False)


print("Data Cleaning Complete. Rows retained:", len(pdf))

[Stage 2:>                                                          (0 + 1) / 1]                                                                                [Stage 3:>                                                          (0 + 1) / 1]                                                                                [Stage 4:>                                                          (0 + 1) / 1]                                                                                [Stage 5:>                                                          (0 + 1) / 1]                                                                                

Medians: 87295.0 130042.0 115024.0
Data Cleaning Complete. Rows retained: 72498

2 2. Salary Distribution by Industry and Employment Type

3 Salary Distribution by Employment Type

import pandas as pd


pdf = df.filter(df["SALARY"] > 0).select("EMPLOYMENT_TYPE_NAME", "SALARY").toPandas()
import re
pdf["EMPLOYMENT_TYPE_NAME"] = pdf["EMPLOYMENT_TYPE_NAME"].apply(
    lambda x: re.sub(r"[^\x00-\x7F]+", "", str(x)).strip() if pd.notnull(x) else ""
)
pdf["SALARY"] = pdf["SALARY"] / 1000
pdf = pdf[pdf["EMPLOYMENT_TYPE_NAME"] != ""]

median_salaries = pdf.groupby("EMPLOYMENT_TYPE_NAME")["SALARY"].median()
median_salaries.head()

sorted_employment_types = median_salaries.sort_values(ascending=False).index

pdf["EMPLOYMENT_TYPE_NAME"] = pd.Categorical(
    pdf["EMPLOYMENT_TYPE_NAME"],
    categories=sorted_employment_types,
    ordered=True
)
fig = px.box(
    pdf,
    x = "EMPLOYMENT_TYPE_NAME",
    y = "SALARY",
    title = "Salary Distribution by Employment Type",
    color_discrete_sequence=["black"],
    boxmode = "group",
    points = "all",
)

fig.update_layout(
    title=dict(
        text="Salary Distribution by Employment Type",
        font=dict(size=16, family="Helvetica", color="black", weight = "bold")
    ),
    
    xaxis=dict(
        title=dict(text="Employment Type", font=dict(size=14, family = "Helvetica", color = "black", weight = "bold")),
        tickangle=0, 
        tickfont=dict(size=12, family = "Arial", color = "black", weight = "bold"),
        showline = True,
        linewidth=2,
        linecolor="black",
        mirror=True,
        showgrid=False,
        categoryorder="array",
        categoryarray=sorted_employment_types.tolist()
    ),
    yaxis=dict(title=dict(text="Salary (K $)", font=dict(size=14, family = "Helvetica", color = "black", weight = "bold")),
        tickvals=[0, 50000, 100000, 150000, 200000, 250000, 300000, 350000, 400000, 450000, 500000],
        ticktext=["0", "50K", "100K", "150K", "200K", "250K", "300K", "350K", "400K", "450K", "500K"],
        tickfont=dict(size=12, family = "Helvetica", color = "black", weight = "bold"),
        showline = True,
        linewidth=2,
        linecolor="black",
        mirror=True,
        showgrid=False,
        gridcolor="lightgray",
        gridwidth=0.5,
    ),
    font=dict(family="Helvetica", size = 12, color = "black"),
    boxgap=0.7,
    plot_bgcolor="white",
    paper_bgcolor="white",
    showlegend=False,
    height=500,
    width=800,


)



fig.show()
fig.write_html("output/Q1.html")
fig.write_image("output/Q1.svg", width=850, height=500, scale=1)

[Stage 6:>                                                          (0 + 1) / 1]                                                                                

4 Salary Distribution by Industry

import pandas as pd
import re
import plotly.express as px

pdf = df.filter(df["SALARY"] > 0).select("NAICS2_NAME", "SALARY").toPandas()

pdf["NAICS2_NAME"] = pdf["NAICS2_NAME"].apply(
    lambda x: re.sub(r"[^\x00-\x7F]+", "", str(x)).strip() if pd.notnull(x) else ""
)
pdf["SALARY"] = pdf["SALARY"] / 1000
pdf = pdf[pdf["NAICS2_NAME"] != ""]

median_salaries = pdf.groupby("NAICS2_NAME")["SALARY"].median()
sorted_industries = median_salaries.sort_values(ascending=False).index

pdf["NAICS2_NAME"] = pd.Categorical(
    pdf["NAICS2_NAME"],
    categories=sorted_industries,
    ordered=True
)

fig = px.box(
    pdf,
    x="NAICS2_NAME",
    y="SALARY",
    title="Salary Distribution by Industry",
    color_discrete_sequence=["#EF553B"],  # bright red for boxes
    boxmode="group",
    points="all",
)

fig.update_layout(
    title=dict(
        text="Salary Distribution by Industry",
        font=dict(size=30, family="Arial", color="#990000", weight="bold")  # dark red title
    ),
    xaxis=dict(
        title=dict(text="Industry", font=dict(size=14, family="Arial", color="#B22222", weight="bold")),  # firebrick red
        tickangle=45,
        tickfont=dict(size=12, family="Arial", color="#B22222", weight="bold"),
        showline=True,
        linewidth=2,
        linecolor="#B22222",
        mirror=True,
        showgrid=False,
        categoryorder="array",
        categoryarray=sorted_industries.tolist()
    ),
    yaxis=dict(
        title=dict(text="Salary (K $)", font=dict(size=14, family="Arial", color="#800000", weight="bold")),  # maroon
        tickvals=[100000, 200000, 300000, 400000, 500000],
        ticktext=["100K", "200K", "300K", "400K", "500K"],
        tickfont=dict(size=12, family="Arial", color="#800000", weight="bold"),
        showline=True,
        linewidth=2,
        linecolor="#800000",
        mirror=True,
        showgrid=False,
        gridcolor="#F5B7B1",
        gridwidth=0.5,
    ),
    font=dict(family="Arial", size=12, color="#800000"),
    boxgap=0.7,
    plot_bgcolor="#FFF0F0",
    paper_bgcolor="#FFF5F5",
    showlegend=False,
    height=900,
    width=1100,
)

fig.show()
fig.write_html("output/Q2.html")
fig.write_image("output/Q2.svg", width=1100, height=900, scale=1)

[Stage 7:>                                                          (0 + 1) / 1]                                                                                

5 Salary Analysis by ONET Occupation Type (Bubble Chart)

df.createOrReplaceTempView("Job_Postings")

25/09/23 23:55:10 WARN SparkStringUtils: Truncated the string representation of a plan since it was too large. This behavior can be adjusted by setting 'spark.sql.debug.maxToStringFields'.

salary_analysis = spark.sql("""
    Select
        TITLE_NAME AS ONET_NAME,
        PERCENTILE(SALARY, 0.5) AS Median_Salary,
        COUNT(*) AS Job_Postings
    FROM Job_Postings
    GROUP BY TITLE_NAME
    ORDER BY Job_Postings DESC
    LIMIT 10
""")

salary_pd = salary_analysis.toPandas()
salary_pd.head()

import plotly.express as px

fig = px.scatter(
    salary_pd,
    x = "ONET_NAME",
    y="Median_Salary",
    size="Job_Postings",
    title="Median Salary by ONET Occupation Type (Bubble Chart)",
    labels = {"ONET_NAME": "ONET Occupation", "Median_Salary": "Median Salary", "Job_Postings": "NUmber of Job Postings"
    },
    hover_name = "ONET_NAME",
    size_max = 60,
    width=1000,
    height=600,
    color="Job_Postings",
    color_continuous_scale="Plasma"
)
fig.update_layout(
  font_family="Arial",
  font_size = 14,
  title_font_size=25,
  xaxis_title = "ONET Occupation",
  yaxis_title= "Median Salary",
  plot_bgcolor="white",
  xaxis=dict(
    tickangle=-45,
    showline=True,
    linecolor="black"
  ),
  yaxis=dict(
    showline=True,
    linecolor="black"
  )
)


fig.show()
fig.write_html("output/Q3.html")
fig.write_image("output/Q3.svg", width=1100, height=900, scale=1)

[Stage 8:>                                                          (0 + 1) / 1]